Design, Development, and Validation of Driving Simulators for Enhancing the Safety and Sustainability of Electric Microvehicles

Abstract

Micromobility vehicles, e-scooters and e–bicycles in particular, gain an increasing popularity but also receive criticism, mainly due to road safety issues and their carbon footprint, particularly in relation to their Li-ion batteries. Available field data are not sufficient to explore those issues. Important input variables, such as riders’ reaction time, the impact of human factors on riders’ safety, battery performance degradation with time, remain unknown. This paper presents the design, development, initial calibration and validation of two novel driving simulators, one for an e-scooter and one, for an e-bicycle. The simulators are already operational and used to acquire new knowledge on driving behavior and battery performance. By enabling a better understanding of e-vehicle performance and safety, these simulators contribute to reducing the environmental impact of micromobility by optimizing battery usage and improving vehicle design for sustainability. The paper describes the overall configuration and the main technical specifications of both simulators and provides a thorough description of all their mechanical and electromechanical components. It documents the initial calibration process before launching the experiments and presents the validation methodology with the participation of over 100 users. The outcomes of future experiments are expected to be beneficial to (i) researchers who will gain new insights on e-vehicle performance, (ii) users, enabling them to make informed decisions on vehicle choice and riding patterns, (iii) urban planners on improving urban infrastructure design, (iv) vehicle manufacturers on identifying customer needs and enhancing vehicle design for sustainability, and (v) Public Authorities on adjusting vehicle and infrastructure specifications to reduce the carbon footprint of urban mobility.

1. Introduction

Micromobility has gained increased popularity over the past five years. Microvehicles have emerged as a promising solution to urban mobility problems [1,2] and have been established as an important element of most major cities’ transportation systems. They provide rapid door-to-door connections within densely populated areas and take up limited space for both circulation and parking. They can also be used for last mile connections to major transit hubs and stations. E-scooters in particular are easy to ride and playful and represent a low-cost travel option compared to taxis [3]. Electric propulsion makes their use possible for more users and usages and over longer distances compared to conventional bicycles that require important physical effort. They are used for commuting and leisure activities, including tourism and sight-seeing in touristic areas. In addition, they are environmentally friendly when they substitute for private cars. They have also proven to be rather resilient to crises, as they continued operations under the COVID pandemic. Many researchers report that public transit riders shifted towards bicycles e-scooters to avoid on-board congestion and to reduce transmission risk [4].

The growing popularity of micromobility has also received criticism from several researchers and citizens. A first major issue concerns the safety of riders, as many accidents occur involving microvehicles and private cars. In those cases, microvehicle riders suffer from serious injuries, as they are more vulnerable. However, accidents also occur in pedestrianized areas, shared spaces, and sidewalks, mainly due to conflicts between e-scooters and pedestrians. In the latter cases, the most vulnerable road user is the pedestrian, who suffers the most severe consequences. Progressively, in many countries, a legal framework was established to regulate e-scooter usage and free-floating fleets’ operations in general. The different legislations define rider rights and obligations concerning issues such as helmet use, minimum age of user, maximal speed, use of a mobile phone, vehicle equipment, and authorization to use urban infrastructure (sidewalks, reserved lanes, etc.). The relevant legislation is rather intuitive or based on research and data related exclusively to bicycles, as evidence on e-scooters and e-bicycles is scarce. Certain aspects are still overlooked, such as the legal alcohol limit, while important input is missing on driving behaviors, such as reaction times. However, this input is essential in order to make informed decisions on their rights of usage, as well as to design soft mobility infrastructure suitable for e-scooter and e-bicycle usage.

Electric microvehicles have also been criticized for their carbon footprint, as they are less environmentally friendly compared to what was initially believed. First, their Li-ion battery uses lithium, which is a rare metal. Then, most vehicles are transported from China to western markets using means of transportation that are not necessarily eco-friendly. During the usage stage, e-microvehicles may replace walking or cycling and, thus, increase CO₂ emissions. In addition, no official recommendations exist regarding optimal charging/discharging cycles and battery maintenance conditions. Inappropriate usage degrades the batteries’ performance and shortens their lifespan. Certain sources report that the lifespan of free-floating e-scooters in dense areas may be as short as 28–30 days [1]. At the same time, the number of incidents of e-scooter fires and battery explosions have grown worldwide [5,6]. Last but not least, the end-of-life of e-scooters is far from being efficiently resolved. The user cannot easily remove the battery from the vehicle, no accredited repair shops exist to undertake the task, and recycling schemes are extremely few in number.

Overall, electric microvehicles represent a promising solution to urban mobility, but two major aspects need to be further studied and regulated: (i) road safety and (ii) battery management. Otherwise, cities may not benefit from the full potential of these emerging modes. The City of Paris, France, recently decided even to entirely ban free-floating e-scooters due to safety and environmental concerns (April 2023). It should be noted that Paris was the showcase of micromobility, counting over 15,000 shared e-scooters. Consequently, both the users and decision-makers should gain knowledge on influential parameters and on the magnitude of the impact of each parameter in order to make informed decisions on vehicle and battery choice, maintenance and usage practices, urban infrastructure design, and traffic regulation. The data needed for this analysis are not directly observable in the field and can be difficulty obtained through travel surveys. Alternative data sources involving laboratory experiments could be used to test different use cases and scenarios and explore risk in a systemic way including road accident and climate change risks.

In all disciplines, simulators are widely used to reproduce actual events and processes under test conditions [7]. They provide a safe and low-cost testing environment compared to real-world observations that are scarce, particularly in the case of extreme events. Of course, simulators cannot entirely reproduce the complexity of physical or social interactions. The stochasticity of natural systems cannot be entirely described by mathematical formulations. However, the ability of simulators to provide rigorous answers to specific problems is proven by many researchers. They are particularly used for future projections and the reproduction of rare and extreme conditions. The output is used for the definition of proactive measures, as well as to strengthen the resilience of systems in terms of infrastructure and personnel preparedness.

In transportation research, simulators are extensively used in many cases, from traffic network loading to behavioral modeling. Most importantly, simulators are used for safety purposes in order to test risk scenarios without putting in danger the users, the vehicles, and the infrastructure. In such cases, simulators provide crucial evidence that is difficult to obtain otherwise (i.e., by observation or experiments). First and foremost, driving simulators are a valuable research tool to investigate system performance and specific subsystem interactions involving human intervention. Human factors play an important role, as they are involved in over 60% or even 90% of accident occurrences [8,9]. Driving behavior shows important heterogeneity across individuals, but may also vary for the same individual at different contexts (professional or leisure trips) and states (level of fatigue, distraction, etc.) [10]. Secondly, simulators provide valuable information to industrial stakeholders. For example, vehicle manufacturers may enhance the man–machine interface or the battery performance based on actual usage patterns such as speed and acceleration profiles. Thirdly, driving simulators are widely used for professional driver training.

Rail, plane, truck, and car driving simulators are numerous worldwide and are being used by transport operators, research laboratories, and industrial manufacturers. On the contrary, it is not common to design, develop, and operate driving simulators for soft mobility and emerging modes. Bicycle simulators can be found in certain research laboratories [11,12,13,14,15,16,17,18]. Some researchers have adapted Virtual Reality (VR) simulators to evaluate cyclists’ behavior under different circumstances and conditions [11,12,13], while others have developed a VR rig in order to improve interaction and realism [14,15,16,17,18]. Research undertaken using bicycle simulators shows that simulators can be used to investigate and evaluate drivers’ behavior with high accuracy. Studies carried out at University of Delft University of Technology (TU Delft) [17], Oregon State University (OSU) [19], and the University of Missouri [14] adopt a projector or set of displays to project the Virtual Environment image and control the Field of View (FoV) by the simulator distance from the projected wall or display. The use of VR headset is a more mainstream way to display the VR environment to the users. To monitor the steering wheel rotation and the back wheel speed, a rotary encoder or a high-precision potentiometer is used. The University of Missouri adopted a laser and a set of cameras to optically observe and measure the rotation angle of the steering wheel. Moreover, most simulators provide slope simulation and a few the lean feature so as to cover 2 Degrees of Freedom (DoF), but the Korea Advanced Institute of Science & Technology (KAIST) interactive bicycle simulator in Korea [20] adopts a 6 DoF Stewart platform and also implements the air flow distance by using a fan in front of the bicycle driver. The fan intensity increases in relation to bicycle speed. Finally, certain simulators, like in Delft University of Technology (TU Delft) [17], have the ability to adjust the seat and steering wheel position.

Moreover, only a few studies on bicycle simulators provide details concerning the simulators’ validation [21,22,23,24]. The validation process typically involves the participation of up to 35 individuals, most of whom take part in field experiments. The validation criteria for electromechanical efficiency are based on crucial simulator characteristics such as velocity, leaning, steering, and slope performance, followed by user experience evaluation. Additionally, all of these studies have introduced major or minor interventions in the bicycles themselves.

The TAMS simulators, on the other hand, have been developed to avoid any interventions on the vehicle’s body, allowing for direct use in real environments by simply detaching the vehicle from the simulator. Furthermore, the validation process considers the mechanical components, sensors, user feedback, and the Unreal Engine physics features to ensure that the VR vehicle is fully kinematic and interacts seamlessly with the virtual environment and other actors. More details are provided in Section 3, and Appendix A includes a comparison table of the reviewed bicycle simulators and the one developed at the TAMS laboratory.

E-scooters and e-bicycles are fundamentally different to conventional bicycles, as the rider does not need to apply torque on the wheel for rotation. This is achieved by the electric motor. Additionally, most e-scooter vehicles need some initial human-powered speed to initiate the electric motor’s operation. Maximal speeds may be comparable, but speed profiles are dissimilar to conventional bicycles due to important differences in acceleration and deceleration rates. Furthermore, degrees of freedom and equilibrium, rider’s reaction time, field of view, and eye height can differ significantly. Those parameters, however, are extremely important in road design and safety analyses. In the case of private automobiles, the same parameters are used to estimate geometrical and operational features, such as the minimal length of alignments or of no-passing zones. To the best of the authors’ knowledge, no such simulator currently exists for e-scooters and e-bicycles.

The objective of this paper is to present the design, development, calibration, and validation of two novel driving simulators, an e-scooter driving simulator and an e-bicycle simulator, that are purposed to work together in a multiplayer experiment implementation. Both simulators were developed and validated at the Transportation and Ambient Mobility Systems (TAMS) at the University of Patras in Greece. They are currently operating and used for different research purposes mainly related to the impact of human factors, such as alcohol consumption or fatigue, on driving. This research is innovative, as it concerns the development of a novel research tool and can pave the way towards important research findings related, but not limited to, the two major challenges that e-microvehicles currently face, namely, road safety and environmental performance.

The results of future experiments will be of direct interest to electric microvehicle users who will have the necessary research-based data to make informed decisions on vehicle purchase and can gain important knowledge on good and bad practices concerning the performance, lifespan, and explosion risk of Li-ion batteries. Furthermore, urban planners and decision-makers will benefit from evidence-based recommendations on the enhancement of soft mobility infrastructure design. Additional applications of the above research include the review of vehicle specifications to improve safety-related aspects from both the manufacturers’ and the certification authorities’ standpoints. Finally, academia will gain new insight on riding performance and safety analysis, Li-ion battery degradation, and risks.

To conclude, this paper introduces two novel driving micromobility simulators designed to address critical gaps in micromobility research. It presents a comprehensive methodology both for calibration and validation processes involving over 100 participants. The presented simulators enable controlled experimentation on rider behavior and battery performance, offering valuable insights for researchers, urban planners, vehicle manufacturers, and policymakers. By providing a robust platform for data collection, this work contributes to improving road safety, optimizing battery usage, addressing life-cycle assessment, and enhancing micromobility vehicle design and infrastructure planning.

The following sections present materials used and the simulators’ construction method (Section 2), describe the calibration process (Section 3), lay out the validation methodology and protocols (Section 4), and summarize major findings (Section 5).

2. Materials and Methods

This section describes the overall design concept and requirements before providing a detailed description of the mechanical structure and all electromechanical equipment and sensors used.

2.1. Design Concept and Requirements

The development of an efficient motion-based simulator or VR rig requires to carefully handle the integration among hardware, electronics, VR environments and human behavior [25]. Integration is achieved as the user interacts with the vehicle, vehicle sensors interact with the VR environment via a programmable microprocessor (such as ARDUINO), and vice versa. Consequently, the VR environment provides the results on an image display device and the user reacts to visual stimuli. A graphical representation of this cyclic process is given in Figure 1.

Several structural considerations shared by both simulators should be addressed in the design process to build an efficient and safe simulator. The simulators have both common and vehicle-specific monitoring and control designs. For common ones, different types of sensors were used to evaluate their efficiency.

Four basic design requirements were defined for the mechanical structure in order to ensure high simulation efficiency and user safety:

• The structural frame has to be able to support the weight of the driver and the vehicle under all possible operational conditions, including vertical slope variations.
• The structural frame has to be able to support all necessary motors and sensors to record and digitize driving behavior parameters.
• The vehicle body has to remain unaffected in order to be functional when dismounted for other use or replaced by other microvehicles on the simulator.
• The total dimensions have to be kept small in order to fit in the laboratory premises and, even, be portable for demonstrations or experiments in other spaces.

For the needs of the present research, a conventional mid-priced e-scooter model with an 8.5″ wheel was selected in order to be similar to the technical specifications of popular vehicles widely used on the road. The only specific requirement was for the motor to be placed in the rear wheel. This requirement stems from the need to separate the steering angle and speed measurements in both wheels, as well as to increase the adaptability of the equipment to future developments and adjustments to other microvehicles, such as e-bikes. For the e-bicycle simulator, a conventional mid-priced model with a 28″ wheel and 250-watt motor, also placed in the back wheel, was selected for the same reasons as the e-scooter. Turning to the supporting platform, a common way to implement a bicycle simulator is to place an actual bicycle on a bike trainer and adjust the friction gearbox and motors. However, in the case of e-scooters, there is no need to apply and feedback the wheel resistance and friction to the driver’s body, as no physical effort is needed for the movement.

An overview of the VR rig components is provided in Figure 2 and Figure 3 for e-scooter and e-bicycle simulator, respectively. Additional details are given in the following paragraphs, and a comparison table for the developed one and the aforementioned bicycle simulators is presented in Appendix A.

2.2. Implementation

The implementation of simulators involves creating realistic, interactive environments that replicate real-world conditions for testing and evaluation. These simulators incorporate dynamic models of both the mechanical structures and electromechanical components to simulate performance accurately. By integrating hardware and software simulations, they allow for an analysis of stability, efficiency, and user ergonomics while considering sensor reliability and motor performance under various operating conditions.

2.2.1. Mechanical Structure

The mechanical part of the finally developed e-scooter simulator consists of three main components (Figure 2s0):

• A rectangular frame made by a metal squared hollow section (25 mm) with total final dimensions of 250 mm × 1200 mm.
• A double-cylinder roller (50 mm diameter and 180 mm length each) where the e-scooter’s rear wheel is mounted.
• Stabilization parts: (a) a cylinder tube with a diameter of 5 mm and a length of 550 mm that is attached to the backside of the rectangular frame to ensure the stability of the simulator and (b) an X-shaped brace attached both on the e-scooter chassis and the rectangular frame to ensure vehicle stability during operations.
• Concerning the lean function, it has been installed in many bicycle simulators by using different techniques [20,24,26]. Besides Shoman and Imine’s [22] study on simulator sickness evaluation, there are no sufficient data about how the lean feature influences the drivers motion sickness or even more on how it augments the experience realism. Matviienko et al. [27] suggested that using handlebar input induces less simulation sickness than body leaning for lab-static simulators. Taking into account these studies and the fact that the e-scooter driving demands minor-to-no lean behavior (due to wheel radius, vehicle size, and speed limit), a decision was made not to take into account the lean function in the e-scooter simulator prototype in order to avoid user simulator sickness and/or possible injuries by side fall.

The mechanical part of the developed e-bicycle simulator consists of the strongest and heavier materials to support higher bicycle weight, user’s physical effort, and lean function (Figure 3b0):

• A polygonal “V-shaped” main frame made by a metal squared hollow section (40 mm) with total final dimensions of 1000 mm × 1650 mm that holds the bicycle and the bike trainer.
• A triangle base with angled steel blade (40 mm × 40 mm) that supports the main frame with a cross joint and holds the linear actuators (LA), with total final distance of 1000 mm × 980 mm. The cross joint role is dual: on the one hand to offer freedom of movement of 2 degrees (2FoD) without the need of a third actuator, and, on the other hand, to support a large amount of total weight so as to avoid overloading the LA.
• A bike trainer attached to main frame that adjusts the friction gearbox and measures wheel speed.
• Concerning the slope and lean function, two linear actuators (LA) are attached to the triangle base and are connected with the main frame with setscrews and ball joints.

2.2.2. Electromechanical Equipment and Sensors

The design requirements here concern the accurate transmission of movement from the user’s choices, such as turning movements or velocity, to the VR physics engine.

• The steering angle is measured using an incremental rotary encoder (RE) (Figure 2s1) with sensitivity of 400 pulses per revolution. The rotary has two wire signals and, thus, “feeds” the software with 800 steps per revolution. This corresponds to 0.45 degrees sensitivity, which is more than satisfactory for the needs of the present research. The values of RE are mapped to a range of −60° to +60° to reproduce the exact range of e-scooter handlebar rotation.
• Since the e-scooter steering wheel is almost vertical to the ground and the wheel diameter is significantly smaller than the one on bicycles (8″−10″ in the majority of urban models), there is much lower friction in steering rotation. Nevertheless, in order to reproduce the sense of ‘resistance’ when turning the handlebar, a metal spring is attached on both the front wheel and the e-scooter chassis (Figure 2s2). The metal spring has an additional practical value, as it returns the steering wheel to its initial position, i.e., after a riding session, and the simulator gets ready for use without any calibration needed related to starting position.
• The surface slope variations are managed by an LA that is attached to the front part of the rectangular frame (Figure 2s3) offering 1 Degree of Freedom (DoF). The LA includes a built-in potentiometer that provides the absolute position of the actuator. The potentiometer provides values from 0 to 1023 and the stroke extends up to 20 cm; therefore, the LA accuracy is up to 0.2 mm.
• As previously discussed, most e-scooters require an initial speeding up provided by the user’s physical effort. To reproduce this feature, an initial speed “starter” mechanism was installed (Figure 2s4). It consists of a DC motor, a set of gears, and a mechanical switch. One gear is attached to the motor, while the other is attached to one cylinder and acts as a clutch. The switch is used to turn on the motor when the starter is pushed and the two gears are engaged together.
• The wheel speed measurement is achieved by attaching a circular disk filled with 60 holes in its circumference to one cylinder roller. Then, a photoelectric sensor counts the number of passing holes per time interval (Figure 2s5). Since the wheel’s and the cylinder’s circumferences are 67.8 cm and 9.4 cm, respectively, the rotation ratio is approximately 1/4.3. The accuracy depends on the time interval and the rotational speed, but, as a rule of thumb, due to 60 holes on the attached disk, each complete cylinder circle covers 9.4 cm, and therefore each hole represents 0.16 cm (9.4/60).

Similar considerations were made to choose the e-bicycle simulator sensors and electromechanically elements:

• The steering angle is measured using a high-accuracy 5 k ohm potentiometer (Figure 3b1). The potentiometer provides values from 0 to 1023 to ARDUINO in a circular range from 0 to 300 degrees and corresponds to 0.29 degrees sensitivity, which is more than satisfactory for the needs of the present research. The values of potentiometer are mapped to a range of −60° to +60° to reproduce the exact e-bicycle handlebar rotation.
• In order to reproduce the sense of ‘resistance’ when turning the handlebar, a metal torsion spring is attached on the box that covers the potentiometer and holds the front wheel with a metal curved cover (Figure 3b1). The metal spring has an additional practical value, as it returns the steering wheel to its initial position, i.e., after a riding session, and the simulator gets ready for use without any calibration needed related to the starting position.
• The surface slope variations and lateral rotation (lean function) are managed by two linear actuators (LAs) that are attached to the back side of the triangle base and the V-shaped main frame (Figure 3b2) and offer a 2-DoF. Each actuator includes a built-in potentiometer that provides the absolute position of the actuator. The potentiometer provides values from 0 to 1023 and the stroke extends up to 25 cm; therefore, the LA accuracy is up to 0.25 mm.
• The wheel speed measurement is achieved by attaching a rotary encoder (RE) to the bikers’ trainer cylinder with a high magnet and a spring that serves as a regulator to smooth any fluctuation due to limited alignment (Figure 3b3). The encoder offers vibration resistance and sensitivity of 360 pulses per revolution and has two wire signals; thus, it “feeds” the software with 720 steps per revolution. Since the bicycle wheel circumference is about 200 cm and the trainers’ cylinder is 9.4 cm, the rotation ratio is 1/21.2. As stated previously, the accuracy depends on the time interval and the rotational speed. Due to 720 pulses, each complete cylinders’ circle covers 9.4 cm and 720 pulses that considering distance accuracy is up to 0.01 cm (9.4/720).

Figure 4 summarizes the common and vehicle-specific construction characteristics of the two simulators.

2.2.3. Virtual Reality Environment (VRE) Development

VR technology can have several advantages when applied to micromobility, including the following:

• Improving rider safety and training: VR can be used to simulate a number of different scenarios and risk environments, allowing for riders to practice and become more familiar with safe riding techniques before actually using a real vehicle. Also, to train riders on the proper use of micromobility vehicles, such as how to properly operate e-scooters or e-bicycles and safety information about the technology.
• Enhancing rider experience: VR can be used to create immersive and interactive experiences for riders, such as virtual tours of cities or virtual races with friends.
• Testing and design: VR can be used to test and design new micromobility vehicles, allowing for manufacturers to evaluate performance, stability, and other important aspects before going into production.
• Simulating scenarios to address electric vehicle driving challenges: VR can be used to tackle issues specific to electric vehicles, such as range anxiety (the driver’s fear of insufficient battery capacity) or battery malfunctions.
• Cost-effective: VR is cost-effective, as it eliminates the need for physical prototypes and reduces the costs associated with testing and design.

The e-scooter and e-bicycle are positioned so as to face the projector screen of the VR environment. The distance between simulators and projected image has to be fit the Field of View (FOV) of the user’s eye to improve realism and immersion. The first VR environment was developed in order to calibrate and validate the simulators before using them for new experiments. Consequently, the first scenes and scenarios produced in the Unreal Engine (UE) software were representing actual roads, buildings, and objects around the University (Figure 5). Real-world data were obtained through observation and field experiments and were, afterwards, used for comparison. For the calibration process, the complete observation protocol can be found in [28] and refers to e-scooters and e-bicycles interacting with pedestrians within a relatively limited area. The detailed description of the validation process is given is Section 3. For the validation process, the main roads in the city center were used covering diverse geometric configurations and traffic volumes (Figure 6). A detailed description of the validation process is given in Section 4.

3. Calibration Process and Results

The calibration stage involves the processes and minor adjustments needed to better reproduce the real riding experience. It tests and enhances (i) the reliability between the VR rig and virtual vehicle regarding the parameters feedback (speed, steering angle, etc.) and (ii) the user’s experience feedback. Four variables were selected for testing, and a range of values was evaluated for each leading to the final choice of optimal values after iterations and a rating process from participants (

Table 1. Calibration variables summary.

Variable	Values Range	Step	Optimal Value
FoV	80–140	15	95, 125
Wheel torque	100–500	100	300
Brake effect	0.01–0.2	0.05	0.15
Steering resistance	0–10	2	4

).

The four variables were chosen on the grounds of their importance upon the degree of realism of simulated riding. Field of View (FoV) is the angular extent of the observable world that is seen at any given moment. A good vision experience requires FoV to be set according to the distance separating the projected image size and the rider’s eye. Wheel torque and Brake effect are essential functionalities, as they ‘replace’ the inertia of the human body on acceleration and braking. In order to maintain the driving performance behavior in the VR while keeping the interactivity in the VR rig, the physics feature must be enabled in the VR application (so as to embed gravity, collisions, friction, and other real effects). Last but not least, steering resistance is controlled by using a turnbuckle to add tension to the spring in order to meet the desired resistance.

Each variable was associated with a specific simulator parameter that can be perceived and rated by users (

Table 2. Assessment criteria parameters.

Variable	Parameter	Description
FoV	Distance assessment	Distance from buildings, obstacles, etc.
Wheel torque	Sense of speed	How close to field speed
Brake effect	Sense of braking	How close to real vehicle braking
Steering resistance	Handlebar performance	Steering realism
n/a	Overall performance	Realism of the whole experience

). The overall simulator performance was also added in the ratings. Participants tested different combinations of values and were asked to rate those parameters after the riding experience. As a benchmark, they were asked to use the exact same experience (i.e., same vehicle, same road environment, similar interactions to other road users) they had in the field just before using the simulator. Overall N = 112 participants aged from 18 to 28 (43% male, 57% female) undertook the field and the simulator experiment. Readers may refer to [28] for more information on the experiment, including the recruitment process and criteria, the description of the field observation protocol, instrumentation, and traffic conditions.

The simulator experimental protocol consisted of four parts: (i) information and guidelines concerning the simulator operation to avoid self-injury and hardware overload, (ii) familiarization with the simulator by offering 2–3 min driving in a demo level, (iii) 3 min (approximately) of simulated riding at desired speed while avoiding obstacles and respecting traffic regulations (Figure 7), and (iv) answering the rating questionnaire using the Likert scale to evaluate performance: 1 for very low, 2 for low, 3 for medium, 4 for good, and 5 for great.

The following charts (Figure 8, Figure 9, Figure 10 and Figure 11) present the participants preference for each assessment criteria in a boxplot format, so that the min, max, and average value for each variable step are shown for both vehicles. Results concerning FoV indicate a comfort range from 95 to 125 with the 95 and 125 values achieving the higher scores for both simulators (Figure 8). Both values were kept for the validation stage. On the contrary, only one value for both simulators (i.e., 300) was retained for the wheel torque (Figure 9). Regarding the sense of braking, the value of 0.15 seems to be clearly preferred by both users (Figure 10). Turning to handlebar performance, two distinct values were retained, i.e., 6 for e-scooters and 8 for e-bicycles (Figure 11). The overall assessment of the simulated experience included comfort, realism, and sensors’ feedback. Over 70% of the participants rated the performance as ‘good’ or ‘great’, with the e-scooter simulator receiving the highest scores (Figure 12). Markedly, these percentages were further enhanced after the calibration process and the final adoption of the optimized values presented in Section 4.

4. Validation Process and Results

In this section, the protocol and the results for VR rig validation are presented. The main scope of the validation process is to assess the reliability of simulated results by comparison to the ground truth, i.e., to real-life usage of micromobility vehicles. Two variables were used for the assessment: speed and acceleration. The validation protocol involved the participation of micromobility users who drove the exact same routes once in the field (in vivo) and once on the simulator (in vitro) using similar microvehicles and ad hoc built VR environments. Their trajectories were registered and compared. To ensure the generality of results, effort was made to increase the heterogeneity of routes and participant characteristics. The three routes (Figure 6) were 1–2.5 km long with flat and non-flat parts and included pedestrianized streets, cycle paths, and mixed traffic lanes.

Participants (N = 114) were again recruited on a voluntary basis from the university community. Participants having participated to the calibration process were screened out. The sample is relatively gender balanced (F(male) = 45.5%), but very young (ages between 18 and 44) due to increased student participation. However, the introduced bias is not very important, as people using microvehicles mainly come from this age group, at least in Greece. Field measurements were obtained using the Physical Phone Experiments application (phyphox smartphone software v.1.1.12, website: https://phyphox.org/, accessed on 10 December 2023 that was installed on their mobile phones and recorded data every 10 m. The app uses phone sensors to provide acceleration and position. In total, 168 circuits were realized in the field and another 168 in the lab, as certain participants run more than one route.

Figure 13 provides indicative individual measurements obtained for the variables tested, namely speed (in km/h), and Figure 14 for acceleration rate (in m/s²) over the length of the three routes considered (R1, R2, and R3). Simulation results (in blue color) are compared to field measurements (in green color). In general, the results are close and follow the same trends. Greater and systematic error is suspected in the case of pedestrianized street (R1). This could be explained by the increased stochasticity of real urban environments where many ‘unexpected’ events may occur. The VR environment simulated both fixed and mobile obstacles such as pedestrians, cars, and microvehicles, but seems not to fully capture real life contexts. The integration of additional obstacles in the VR environment would improve results and is a future planned improvement of the simulators.

Aggregate comparison results for the total of 168 circuits realized are shown below for speed (Figure 15) and acceleration (Figure 16), and they confirm previous statements. In all cases, the average difference between simulator and field, as measured per individual circuit, is lower than 15%. The highest speed difference is observed on the pedestrianized street (R1) for e-scooters, where the simulator overestimates speed by 13.5%. The mixed traffic lane (R3) shows the lowest error; for the case of e-scooters, it only shows 4.1%. On the contrary, the simulator seems to consistently, in all routes, underestimate acceleration rates, with higher differences being again observed in the case of pedestrianized street R1. The speeds in the virtual environment are quite close to those in reality on all three routes of interest, and the percentage difference ranges from 4.1% to 13.5% (absolute value). As for the overall performance parameter, there is a high improvement from previous evaluation in the calibration process and high-to-great-grade rises up to 94.3% for the e-scooter and up to 91.9% for the e-bicycle (Figure 17), meaning that the calibration process settings step up both simulators performance. Overall, the results are very encouraging regarding the simulators’ ability to reproduce field riding conditions. The maximum error of 10% observed on bicycle paths and mixed traffic lanes allows us to consider the simulator validated. However, certain improvements should be made for pedestrianized streets to reduce the error in the case of acceleration.

5. Conclusions

Micromobility vehicles, e-scooters, and e–bicycles in particular, have gained increasing popularity, but have also received criticism, mainly due to road safety issues and their carbon footprint in relation to their Li-ion batteries. Available field data are not sufficient to explore these issues. Important input variables, such as riders’ reaction time, the impact of human factors on riders’ safety, and battery performance degradation with time, remain unknown. This paper presents the design, development, initial calibration, and validation of two novel driving simulators: one for an e-scooter and one for an e-bicycle.

This paper describes the overall configuration and the main technical specifications of both simulators and provides a thorough description of all their mechanical and electromechanical components. All initial requirements are satisfied, common and vehicle-specific designs are detailed, and the VR environment is presented. Possible future improvements were identified after comparison of the novel simulators to existing conventional bicycle simulators. In particular, the main base body design might accept the simulation of uneven ground vibrations, and the “starter” mechanism may be designed to operate by foot pushing and also may not. In addition, the lean feature may be integrated into the e-scooter simulator in the future.

Afterwards, a structured experimental calibration process allowed for the adjustment of certain elements related to the distance assessment, the sense of speed and braking, and the steering resistance. The experiment was realized with the participation of 112 participants who were also asked to rate the overall performance of the simulated experience. The calibration process led to the definition of adjusted values having obtained higher scores from microvehicle users. After calibration, satisfaction rates increased to over 90%.

A second experimental process was designed and undertaken for validation purposes. The error between simulated results, i.e., speed and acceleration, and field results was measured on three routes, including pedestrianized streets, bicycle paths, and mixed traffic lanes. A double experiment, in vivo and in vitro, was performed with the participation of 114 micromobility users. Results show that the errors are low and acceptable, especially in the case of bicycle and mixed traffic lanes. They are slightly higher on pedestrianized streets, where more interactions and unexpected events may occur. As a result, additional moving obstacles will be added in the VR environment to result in lower speeds that are closer to reality.

Overall, the two novel simulators are validated and are used for further investigations on microvehicle safety and energy consumption. The outcomes of future experiments are expected to be highly beneficial to (i) researchers who will gain new insights on e-vehicle performance, (ii) users, enabling them to make informed decisions on vehicle choice and riding patterns, (iii) urban planners on improving urban infrastructure design, (iv) vehicle manufacturers on identifying customer needs and enhancing vehicle design, and (v) public authorities on adjusting vehicle and infrastructure specifications.

Future research could involve the integration of additional hardware devices, such as haptic feedback systems, which would need to be calibrated with the current simulator to enhance the interaction between the user and the virtual reality environment (VRE). Furthermore, sensors focusing on drivers’ reactions to environmental and traffic incidents—such as eye-tracking hardware, heart rate pulse sensors (Electrocardiogram—ECG), and facial tracking for emotion detection—could be implemented to provide deeper insights into drivers’ behavior. Finally, by implementing the aforementioned hardware updates, future research could focus on addressing range anxiety to study user behavior, energy consumption, and route optimization in a controlled virtual environment. The simulator can help test strategies like real-time range prediction, battery management systems, and charging station placement, providing valuable insights to improve confidence in electric micromobility. These findings could guide the development of user-centered solutions, enhancing the overall experience and encouraging the broader adoption of electric vehicles.